View
17
Download
0
Category
Preview:
Citation preview
Proceedings of ICAPP ‘10 San Diego, CA, USA, June 13-17, 2010
Paper 10141
Interaction of Computational Tools for Multiscale Multiphysics
Simulation of Generation-IV Reactors
I.A. Bolotnov1, F. Behafarid
1, D. Shaver
1, T. Guo
2, S. Wang
2, H. Wei
2,
S.P. Antal1, K.E. Jansen
1, R. Samulyak
2,3, M.Z. Podowski
1*
1Center for Multiphase Research, Rensselaer Polytechnic Institute;
2Department of Applied Mathematics and Statistics, Stony Brook University
3Computational Science Center, Brookhaven National Laboratory
110 8th
St., Troy, NY, U.S.A. *Tel: (518)276-4000, Email:podowm@rpi.edu
Abstract – The objective of this paper is to discuss recent accomplishments in the development of
advanced computer simulation capabilities for safety analyses of next-generation nuclear reactors.
The proposed approach combines high performance computing with the necessary level of modeling
detail and high accuracy of predictions. The results of multiscale simulations are shown of an accident
scenario in a sodium-fast reactor, obtained using three interacting fluid dynamics codes, each
performing simulations at a different scale: FronTier, PHASTA and NPHASE-CMFD. The loss-of-flow
accident scenario results in the overheat and breach of the cladding of a stainless steel fuel, which
results in the injection of pressurized fission gas into the molten sodium coolant.
1. INTRODUCTION
It has been widely recognized that the development of
next generation nuclear reactors, in particular - Gen. IV
reactors, will require modeling and simulation tools which
are much more accurate than those used so far by the
nuclear industry, and are also capable of capturing a broad
spectrum of multiscale physical phenomena governing
reactor transients and accidents in a mechanistic manner.
The development of a suite of high performance
computational tools for multiscale simulations of the Gen-
IV reactors has recently been undertaken by a DOE-
sponsored university consortium. An important aspect of
this project is concerned with simulating the phenomena
leading to, and governing the progression of, core blockage
accidents in the proposed Gen. IV Sodium Fast Reactors
(SFR).
The purpose of this paper is to present a 3-D modeling
concept and its multi-code computer implementation, to
simulate the injection of a jet of gaseous fission products
into a partially blocked SFR coolant channel following
localized cladding overheat and breach. Three computer
codes have been used in the analysis, namely: FronTier,
PHASTA and NPHASE-CMFD.
The FronTier code has been used to model the heatup
and melting of a nuclear fuel rod, failure of the rod‟s
stainless steel cladding, and the escape of fission gases
through the crack in the cladding. The output of FronTier
simulations has been coupled with a fluid mechanics direct
numerical simulation (DNS) code, PHASTA.
The PHASTA code was coupled with a Level-Set
model and used to simulate the injection of a two-phase
fission gas jet into the coolant channel filled with liquid
sodium. The time-dependent inflow shape and velocity
profile of the fission gas is obtained from the FronTier
results. The instantaneous velocity and gas concentration
profiles were used to collect the necessary statistics
downstream of the injection point, to provide multiphase
inflow conditions for the Reynolds averaged Navier Stokes
(RaNS) equations multiphase flow solver, NPHASE-
CMFD.
A three-dimensional model of gas/liquid-sodium
interaction has been developed based on a multifield
modeling framework implemented in the NPHASE-CMFD
code. The NPHASE-based simulations were performed for
a computational domain which was comparable in size with
the geometry of complete interconnected reactor coolant
channels.
Details of the current multiscale multiple-solver
modeling concept, as well as the results of predictions for a
hypothetical channel blockage accident in the proposed
sodium fast reactor are discussed below.
2. DESCRIPTION OF INDIVIDUAL COMPUTER
CODES
A. FronTier
FronTier is a computational package for direct
numerical simulation of multiphase flows [1] developed at
Stony Brook University in collaboration with LANL and
BNL. FronTier development and optimization for
massively parallel supercomputers is a part of the DOE
SciDAC program. An important and unique feature of this
package is its robust ability to track dynamically moving
fronts or material interfaces using the method of front
Proceedings of ICAPP ‘10 San Diego, CA, USA, June 13-17, 2010
Paper 10141
2
tracking [2]. FronTier supports compressible and
incompressible Navier-Stokes equations and MHD
equations in the low magnetic Reynolds number
approximation [3], and phase transitions such as melting
and vaporization [4]. The FronTier code has been used for
large scale simulations of a variety of physical problems [3,
5, 6] on various platforms including the IBM BlueGene
supercomputer New York Blue located at Brookhaven
National Laboratory. The application of FronTier to the
present problem has been two-fold. First, it has been used
to model solid-state mechanics phenomena governing
cladding heatup and breach. Secondly, using the initial
conditions of pressurized fission gas inside the reactor fuel
pins prior to cladding failure, simulations have been
performed of gas flow through the cracked cladding.
B. PHASTA
PHASTA is a parallel, hierarchic (higher order
accurate from 2nd-5th order accurate depending on
function choice), unstructured/adaptive, stabilized (finite
element) transient analysis flow solver (both
incompressible and compressible) [7, 8]. PHASTA (and its
predecessor, ENSA) was the first unstructured grid LES
code [9] and has been applied to turbulent flows ranging
from validation benchmarks (channel flow, decay of
isotropic turbulence) to complex flows (airfoils at
maximum lift, flow over a cavity, near lip jet engine flows
and fin-tube heat exchangers). The PHASTA code uses
advanced anisotropic adaptive algorithms [10] and the most
advanced LES/DES models [11]. Note that DES, LES, and
DNS are computationally intensive even for single phase
flows. The two-phase version of PHASTA utilizes the
Level Set method to define the interface between the gas
and liquid phases [12]. Here, it has been used to simulate
gas jet injection into the coolant region using the boundary
conditions provided by FronTier. Turbulent two phase
PHASTA outflow is averaged over time to obtain mean
velocities for each phase, turbulent kinetic energy and
turbulence dissipation rate of each phase. This time-
averaged data provides the input to the next scale of the
simulation using the NPHASE-CMFD code. A sliding
window time averaging has been used to capture slow
changes to the mean flow parameters. Also, PHASTA
provides the pressure found at the inflow of its domain
back to FronTier outflow
C. NPHASE-CMFD
NPHASE-CMFD is an advanced Computational
Multiphase Fluid Dynamics computer program [14] for the
simulation and prediction of combined mass, momentum
and energy transfer processes in a variety of single-phase
[15] and multiphase/multiscale systems, including
gas/liquid [16, 17], solid/liquid [18, 19] and
gas/solid/liquid [14] flows. It uses two-phase k-ε models of
turbulence (the user can chose between the High Reynolds
Number and Low Reynolds Number options of the model).
In the present work, transient inflow boundary conditions
have been supplied by PHASTA to simulate fission-
gas/liquid sodium two-phase flow along and around the
failed fuel element and the elements adjacent to it.
NPHASE-CMFD also supplies the pressure value back to
the PHASTA domain outflow. Thus, all the three codes
used for the described multifield simulation are fully
coupled.
3. MODEL DESCRIPTION
3.1. Physical Problem
As it was mentioned before, a hypothetical channel-
blockage accident scenario in Gen-IV SFR has been used
as testing ground for the proposed approach of using three
interacting codes in multiscale simulations. Fig. 1 presents
an overview of the problem geometry. It has been assumed
that as a result of the accident, one of the fuel rods
overheats and causes a failure of the stainless steel
cladding. This results in fission gas injection into the
coolant.
Fig. 1. Geometry and dimensions of a section of SFR fuel
rod assembly with the cladding failure location.
The modeled phenomena include:
stainless steel failure mechanisms
pressurized fission gas escape from the fuel rod into
the liquid sodium coolant
two-phase side jet injection into the coolant
transient two-phase flow propagation downstream
from cladding failure
d = 10.4 mm
L ~
1.0
m
Overheated fuel rod
R = 4.0 mm
Proceedings of ICAPP ‘10 San Diego, CA, USA, June 13-17, 2010
Paper 10141
3
3.2. Computational Domains
Figures 2 and 3 show the computational domain
sharing in the multiscale approach used in the simulations.
A description of the domain used by each of the three codes
and their interfaces in given next.
Fig. 2. Computational domains overview (side view).
Fig. 3. Top view of the computational domains for FronTier
and PHASTA.
FronTier computational domain
FronTier resolves detailed physics processes in the
computational domain that includes a 6 cm long section
around the hottest region of the fuel rod and a 1 cm thick
layer of the sodium coolant adjacent to the rod.
Computational grid use in the FronTier simulations of
cladding breach is shown in Fig. 4.
Fig. 4. Computational mesh in the FronTier simulations of
cladding breach.
PHASTA computational domain
The PHASTA domain is only about 1 cm tall, but it
includes the region around the cladding breach, which
overlaps with the FronTier domain, as well as section of the
coolant channel which starts below the breach and extends
to a short distance above the breach.
The finite element mesh used in PHASTA simulation
is shown in Fig. 5. It is characterized by the following
parameters:
Number of vertices: 2.145 million
Number of elements: 12.13 million
The gas flow rate computed by FronTier simulations is
used for setting up gas inflow boundary conditions in
PHASTA. Other boundary conditions are: no-slip boundary
conditions applied on all solid surfaces, and a pressure
boundary condition that corrects for the effects of surface
tension at the outflow of the domain. This outflow pressure
is supplied by the NPHASE flow solver.
NPHASE computational domain
The computational domain of NPHASE, shown in Fig.
2, actually encompasses and extends along, the coolant
channels around 30 fuel rods. This is shown in Fig. 6,
where the NPHASE domain overlays the PHASTA domain.
x
z
Overheated
Fuel Rod
Neighbor
Fuel Rod
Coolant
Inflow
NPHASE
Domain
PHASTA/NPHASE Interface
FronTier/PH
ASTA
interface PHASTA
Domain
Fission gas
FronTier
Domain
PHASTA
inflow
PHASTA
domain
y
z
FronTier
domain y
z
Proceedings of ICAPP ‘10 San Diego, CA, USA, June 13-17, 2010
Paper 10141
4
Due to the symmetry of the problem and the ensemble
averaged nature of the NPHASE calculation, the data taken
from PHASTA is averaged across the symmetry line shown
in Fig. 6. The corresponding computational mesh in the
NPHASE simulations is shown in Fig. 7.
Fig. 5. Computational mesh used in PHASTA simulations.
Fig. 6. Cross section of the NPHASE domain showing
overlap with the PHASTA domain (blue) and the added
reactor channels (red). Arrow indicates the location of gas
jet. Solid line indicates the location of plots presented in
Fig. 19.
The mesh used in the NPHASE simulation is shown in
Fig. 7. It is axially non-uniform and consists of
approximately 500k elements. Since the NPHASE mesh is
much coarser than the PHASTA mesh, data is interpolated
from the PHASTA onto the NPHASE meshes to generate
the inlet conditions for the NPHASE simulation. The total
physical length of the NPHASE domain is 0.5m. This, in
turn, illustrates the advantages of coupling the DNS results
of PHASTA, limited for practical reasons to a relatively
small computational domain, with the NPHASE-based
large-scale model which uses accurate PHASTA
predictions as input.
Fig. 7. Computational mesh used in the NPHASE
simulations
3.3. Description of mathematical models
FronTier Models
The FronTier package has the ability to evaluate
moving fronts or material interfaces using the method of
front tracking. As shown in Fig. 8, the interfaces are
represented as lower dimensional meshes moving through a
volume filling grid. A volume-filling finite difference grid
supports smooth solutions located in the region between
interfaces. The dynamics of the interface comes from the
mathematical theory of the Riemann problem [20].
Fig. 8. Rectangular grid, interface, and states for the
method of front tracking. States contain density,
momentum, and energy density of the fluid, and references
to the EOS model and other parameters.
In the FronTier's solution algorithm [21], the Riemann
problem is solved for the left and right interface states to
predict the location and states of the interface at the next
time step. Then a corrector technique is employed which
accounts for fluid gradients on both sides of the interface.
FronTier can evolve and resolve topological changes of
a large number of interfaces in 2D and 3D spaces.
Proceedings of ICAPP ‘10 San Diego, CA, USA, June 13-17, 2010
Paper 10141
5
Additional features of the FronTier hyperbolic solvers
include a choice of either exact or approximate Riemann
solvers and realistic models for the equation of state.
For the simulation of overheating and melting of the
nuclear fuel rod, new elliptic/parabolic solvers have been
developed, based on the embedded boundary method [4],
as well as an algorithm for the classical Stefan problem
describing first order phase transitions.
A new approach, based on the energy minimization of
the network of springs with critical tension, has also been
applied to develop new mesoscale models for the
simulation of fracture of inhomogeneous solid materials.
Such a description of solids makes it possible to avoid fast
time scales associated with acoustic waves and capture
important features of the material fracture. The main
emphasis has been on the study of brittle fracture. Two
regimes of the fracture have been considered: (a)
adiabatically slow deformation and breakup and (b)
instantaneously fast deformation and the formation and
propagation of cracks in stressed materials. The most
computationally intensive step in the algorithm is the
energy minimization. The methodology applied to both
constrained and unconstrained optimization problems is
done through the interface with TAO, a package developed
by Argonne National Lab within the DOE SciDAC
initiative.
Model testing included a comparison of the breakdown
stress with experimental data, studies of the breakdown
stress as a function of the probability of defects up to the
rigidity percolation threshold, and scaling of the
distribution function of the breakdown stress.
PHASTA models
The spatial and temporal discretization of the
Incompressible Navier-Stokes equations within PHASTA
has been described in Whiting and Jansen [7]. The strong
form of the INS equations is given by
Mass conservation
, 0i ju (1)
Momentum conservation
, , , ,ˆˆ ˆ
i t j i j i ij j iu u u p f (2)
where ρ is density, ui is the i-th component of velocity,
ˆ,p i is pressure, ˆ
ij is the stress tensor, and f̂i represents
body forces along the i-th coordinate.
For the incompressible flow of a Newtonian fluid, the
stress tensor is related to the strain rate tensor, Sij, as
, ,ˆ 2 ( )ij ij i j j iS u u (3)
Using the Continuum Surface Tension (CST) model of
Brackbill et al. [22], the surface tension force is computed
as a local interfacial force density, which is included in f̂i .
The stabilized finite element formulation used in the
PHASTA code [7] is based on the work of Taylor et al.
[23], Franca and Frey [24].
The level set method [12, 25, 26, 27] involves
modeling the interface as the zero level set of a smooth
function, φ, where φ represents the signed distance from the
interface. Hence, the interface is defined by φ = 0. The
scalar φ, the so-called first scalar, is convected within a
moving fluid according to,
. 0D
uDt t
(4)
where u is the flow velocity vector. Phase-1, typically the
liquid phase, is indicated by a positive level set, φ > 0, and
phase-2 by a negative level set, φ < 0. Evaluating the jump
in physical properties using a step change across the
interface leads to poor computational results. Therefore, the
properties near an interface are defined using a smoothed
Heaviside kernel function, Hε, given by:
,0
1 1 ,( ) 1 sin2
,1
H
(5)
The fluid properties are thus defined as
1 2( ) ( ) (1 ( ))H H (6)
1 2( ) ( ) (1 ( ))H H (7)
Although the solution may be reasonably good in the
immediate vicinity of the interface, the distance field may
not be correct throughout the domain since the varying
fluid velocities throughout the flow field distorts the level
set contours. Thus the level set is corrected with a re-
distancing operation [27].
NPHASE physical model
The multiphase model in the NPHASE-CMFD code is
based on the multifield modeling concept. The ensemble-
averaged mass and momentum equations, respectively, are
given by
Mass conservation
k k
k k k kt
v
(8)
Proceedings of ICAPP ‘10 San Diego, CA, USA, June 13-17, 2010
Paper 10141
6
Momentum conservation
( )
( )
k k k
k k k k
i t
k k k kj k k k
j
t i i
k kj k kj k k
j j
t
p p p
vv v
M g
(9)
In Eqs.(8)-(9), (j,k) are the field indices, the subscript
„i‟ refers to the interfacial parameters,
i
kjM is the interfacial
force per unit volume exerted by field-j on field-k, and the
remaining notation is conventional.
The total interfacial force is formulated as a
superposition of several component forces
i D VM L TD
kj kj kj kj kj M M M M M (10)
where D
kjM is the drag force, is the VM
kjM virtual mass force,
L
kjM is the lift force, and TD
kjM is the turbulent dispersion
force.
The turbulence model in the NPHASE-CMFD code
accounts for both the continuous-field sheer-induced
turbulent viscosity and the dispersed-field(s)-induced
viscosity
turb SI BI
c c c (11)
The continuous field sheer-induced turbulence is
modeled using one of two versions of the k-e model: High
Reynolds Number (HRN) or Low Reynolds Number (LRN)
models. The dispersed-phase induced turbulence is based
on the model proposed by Sato & Sekoguchi [28].
3.4. Boundary conditions and interfaces between the
codes
To achieve coupling between the three codes used in
the analysis, a consistent data transfer must be provided
between the individual codes. A file-based data transfer
method has been used to provide multiphase inflow
boundary conditions for “downstream” software, using
additional processing, such as space and time interpolation
as well as averaging techniques when converting the DNS
results into RaNS inflow boundary conditions.
FronTier to PHASTA interface description
The FronTier crack formation simulation generates
mesh coordinates along with a connectivity array for the
crack shape at each time step of the crack growth
simulation. The PHASTA simulation uses this information
to block selected internal mesh nodes in its computational
domain, to dynamically change the effective crack size in
the two-phase flow simulation.
Since both FronTier and PHASTA provide
instantaneous velocity fields as their solutions, no
averaging of the data is required for this interface. The
following steps are taken to use the fission gas velocity
profile flowing though the failed stainless steel cladding
opening computed by FronTier as PHASTA inflow
boundary conditions:
Velocity distribution is recorded by FronTier in the
plane of interests
For each FronTier time step velocity profiles are
interpolated on the PHASTA mesh inflow nodes
During the PHASTA run the time-dependent velocity
profiles from FronTier are used as jet inflow
boundary conditions (linear time interpolation is
performed when PHASTA time step is smaller than
FronTier time step)
PHASTA to NPHASE-CMFD data transfer
The DNS results produced by the PHASTA code
require averaging before they can be used in NPHASE-
CMFD as inflow boundary conditions. The following
technique is used to provide the boundary conditions for
the NPHASE domain:
A list of coordinates at the NPHASE domain inflow is
generated.
At every time step of PHASTA execution
instantaneous velocity components (u1, u2, u3) and
distance to interface field are interpolated from
PHASTA solution at the NPHASE coordinates.
A postprocessing code performs the analysis of the
data and computes the mean velocity distribution
(U1k, U2
k, U3
k), volume fraction, turbulent kinetic
energy and turbulence dissipation rate, for each flow
field, k (e.g., k = 1for liquid and k = 2 for gas).
An example of velocity fluctuations along with gas phase
indicator function is shown on Fig. 9.
4. RESULTS
The models discussed in Section 3 have been used to
simulate the consequences of a loss-of-flow accident in a
Gen. IV Sodium fast reactor (SFR). The phenomena
modeled included: fuel element heatup, cladding heatup
and failure due to combined effects of thermal stresses and
fission gas pressure inside the fuel pin, injection of
pressurized fission gas into the coolant channels
surrounding the failed fuel element, and the transport of gas
/liquid-sodium mixture toward the top of the core. The
results of the multiscale simulations are presented next.
Proceedings of ICAPP ‘10 San Diego, CA, USA, June 13-17, 2010
Paper 10141
7
-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
ui,m/s
time, s
ui X2
Fig. 9. Example of DNS data produced by PHASTA for
two-phase flow at a fixed point which is used for obtaining
averaged data for NPHASE-CMFD (ui is velocity
components and X2 is the gas phase indicator function;
averaged values 1
1U and 2
1U are shown in blue and orange,
respectively).
4.1. FronTier Calculations
In the initial stage of FronTier simulations, heat
transfer was evaluated inside the hottest section of the fuel
rod with oxidic fuel. The calculated temperature
distribution across the fuel rod at normal operating
conditions is shown in Fig. 10. The predicted temperature
profile reflects the effect of thermal conductivity of both
the fuel and the cladding, as well as of the average thermal
conductance of the gas between the fuel pellets and the
inner cladding wall. The use of the embedded boundary
method for material interfaces was critical for achieving
accurate simulations of the heat transfer problem for such a
system.
Fig. 10. Temperature distribution across the fuel rod.
As a result of a reduced heat removal rate following
loss of coolant flow, the temperatures of both the fuel
material and the cladding wall start increasing. Figure 11
shows the increase of fuel temperature and the resultant
partial fuel melting in the central region of the hottest
pellet.
Fig. 11. Temperature distribution in the fuel rod during
transient overheating and melting of the fuel.
In the present case, the cladding fails before melting,
causing the ejection of fission gases into the coolant
channels. The increasing fuel temperature causes the
pressure inside the fuel rod to increase as well, thus
augmenting the mechanical load on the cladding wall. The
combined effects of elevated gas pressure and cladding
temperature weaken the cladding wall and eventually lead
crack formation. The predicted crack side is shown in 12.
Fig. 12. Formation of crack in the fuel rod cladding.
Proceedings of ICAPP ‘10 San Diego, CA, USA, June 13-17, 2010
Paper 10141
8
In a separate FronTier simulation, the compressible
hydrodynamics front tracking code has been used to
simulate the ejection of fission gases into the sodium
coolant. To obtain the boundary conditions for the gas
pressure in the gap behind the crack, a porous medium
model was used, which resulted in a high pressure drop
between the gas storage volume and the near crack region
inside the fuel pin. The FronTier-calculated pressure
distribution in the crack region is shown in Figure 13.
Fig. 13. Isosurfaces of pressure of the fission gas escaping
the coolant through the crack in the cladding, calculated by
FronTier.
4.2. PHASTA Calculations
The PHASTA code has been used to simulate the
second scale of the multiscale problem. It used the gas
inlet information provided by FronTier. Fig. 14 shows the
gas velocity profile observed in PHASTA simulation inside
the stainless steel cladding failure zone. Fig. 15 shows an
instantaneous velocity profile for a 3D simulation of a gas
jet entering the liquid sodium coolant flow.
Fig. 14. Velocity profile of fission gas injected into the
reactor coolant channels, calculated by PHASTA.
A three-dimensional view of the instantaneous
PHASTA solution is shown in Figs. 15 and 16. The top
surface of the domain shown in Fig. 16 is the
PHASTA/NPHASE interface plane where the flow
statistics are collected in order to provide inflow conditions
for the NPHASE run.
(a)
(b)
Fig. 15. 3D two-phase PHASTA simulation of fission gas
jet entering the coolant flow. This Figure shows both the
velocity magnitude and gas/liquid interface contours: (a)
side view, (b) top view.
Fig. 16. Instantaneous PHASTA velocity field with the
NPHASE interface plane shown at the top. The gas/liquid
interface is marked with a solid black line.
Proceedings of ICAPP ‘10 San Diego, CA, USA, June 13-17, 2010
Paper 10141
9
The present two-phase PHASTA simulation was
performed on 2,048 IBM BlueGene/L processors for about
180 wall clock hours, which corresponds to approximately
370,000 CPU-hours.
Fig. 17. (a) U velocity component of liquid phase; (b) U
velocity component of gas phase; (c) volume fraction of the
gas phase in the NPHASE solution (inlet at right, outlet at
left, equally-spaced intervals in between).
4.3. NPHASE-CMFD Calculations
The NPHASE-CMFD code has been used to simulate
the largest scale of the present multi-scale problem. The
data has been collected at a cross plane in the PHASTA
domain and used at the inlet of the NPHASE simulation.
The inflow data taken from the PHASTA simulation has
been applied to the blue region shown in Fig. 6. For the
inlet condition used in the NPHASE domain in the regions
not covered by the PHASTA simulation, a fully developed
turbulent profile has been used matching the coolant mass
flux specified by PHASTA. This profile has been
calculated using the NPHASE code.
The NPHASE model tracks the velocity fields of the
gas and liquid phases independently. These can be seen in
Figs. 17(a) and 17(b). The complex inlet flow pattern is a
direct result of large scale eddies produced by the jet
hitting the neighboring fuel rods. These are long standing
fluid structures which are not eliminated by the time
averaging done on the PHASTA data. Downstream from
the jet, these structures decay into a much less chaotic flow.
The NPHASE-CMFD model has been used to
determine how the gas released into the coolant spreads as
it travels along the coolant channel. Figs. 17(c) and 18
show how the volume fraction of gas changes along the
reactor channel. The turbulent dispersion force causes it to
spread out over a significant portion of the channel cross
section. Fig. 19 presents the line plots of the volume
fraction of gas taken at various locations along the coolant
flow. It can be seen how the gas redistributes itself before
being dispersed. As can be seen in Fig. 17, the gas velocity
increases along the channel due to gravity. This causes a
drastic drop in volume fraction which is observed in the
initial section of the channel.
Fig.18. 3D computational domain showing the gas
volume fraction predicted by NPHASE-CMFD. The inlet
is at the bottom right, the outlet at top left.
5. SUMMARY AND CONCLUSIONS
A multiscale modeling approach to the analysis of
nuclear reactor transients and accidents has been discussed.
It has been demonstrated that by using multiple computer
codes in a coupled mode, a broad range of phenomena
governing accident progression can be simulated in a
physically-consistent and numerically accurate manner.
Proceedings of ICAPP ‘10 San Diego, CA, USA, June 13-17, 2010
Paper 10141
10
ACKNOWLEDGMENTS
The authors would like to acknowledge the financial
support provide to this study by the U.S. Department of
Energy and the computation resources provided by the
Computational Center for Nanotechnology Innovations
(CCNI) at Rensselaer Polytechnic Institute (RPI).
Fig. 19. NPHASE-CMFD-predicted distribution of gas
volume fraction (plots taken along solid line shown in
Fig. 6) for various streamwise locations given in legend..
REFERENCES
1. Du, J., Fix, B., Glimm, J., Li, X., Li, Y., Wu, L., “A
Simple Package for Front Tracking”, J. Comp. Phys.,
213, 613-628 (2006).
2. Glimm, J., Grove, J., Li, X., Tan, D.C., ”Robust
computational algorithms for dynamic interface
tracking in three dimensions”, SIAM J. Sci. Comput.
21, 2240-2256 (2000).
3. Samulyak, R., Du, J., Glimm, J., Xu, Z., “A numerical
algorithm for MHD of free surface flows at low
magnetic Reynolds numbers:, J. Comp. Phys., 226,
1532-1549, (2007).
4. Wang, S., Samulyak, R., Guo, T., “An embedded
boundary method for parabolic problems with
interfaces and application to multi-material systems
with phase transitions”, Acta Mathematica Scientia, in
press, (2010).
5. George, E., Glimm, J., Li, X., Li, Y., Liu, X.,
“Influence of scale-breaking phenomena on turbulent
mixing rates”, Phys. Rev. E, 73, 0163041-5 (2006).
6. Samulyak, R., Lu, T., Parks, P., “A magneto-
hydrodynamic simulation of pellet ablation in the
electrostatic approximation”, Nuclear Fusion, 47, 103-
118 (2007).
7. Whiting, C.H. and Jansen, K.E., “A Stabilized Finite
Element Formulation For The Incompressible Navier-
Stokes Equations Using A Hierarchical Basis,”
International Journal of Numerical Methods in Fluids,
35, 93-116 (2001).
8. Jansen, K. E., Whiting, C.H. and Hulbert, G.M., “A
generalized-alpha method for integrating the filtered
Navier-Stokes equations with a stabilized finite
element method”, Computer Methods in Applied
Mechanics and Engineering, 190, 3-4, 305-319
(2000).
9. Jansen, K.E., “Unstructured Grid Large Eddy
Simulations of Wall Bounded Flows”, Annual
Research Briefs, Center for Turbulence Research,
NASA Ames/Stanford University, 151 (1993).
10. Sahni, O., Mueller, J., Jansen, K.E., Shephard, M.S.
and Taylor, C.A., “Efficient Anisotropic Adaptive
Discretization of the Cardiovascular System”,
Computer Methods in Applied Mechanics and
Engineering, 195, 41-43, 5634-5655 (2006).
11. Tejada-Martinez, A.E. and Jansen, K.E., “A parameter-
free dynamic subgrid-scale model for large-eddy
simulation”, Computer Methods in Applied Mechanics
and Engineering, 194, No. 9, 1225-1248 (2005).
12. Sethian, J. A., "Level Set Methods and Fast Marching
Methods", Cambridge University Press. (1999).
13. Nagrath, S., Jansen, K.E. and Lahey, Jr., R.T.,
“Computation of incompressible bubble dynamics with
a stabilized finite element level set method”, Computer
Methods in Applied Mechanics and Engineering, 194
(42-44) 4565-4587 (2005).
14. Antal, S., Ettorre, S., Kunz, R. and Podowski, M.,
“Development of a next generation computer code for
the prediction of multicomponent multiphase
flows,” Proceeding of the International Meeting on
Trends in Numerical and Physical Modeling for
Industrial Multiphase Flow, Cargese, France (2000).
15. Gallaway, T., Antal, S.P. and Podowski, M.Z.,
“Multidimensional Model of Fluid Flow and Heat
Transfer in Generation-IV Supercritical Water
Reactors", Nuclear Engineering and Design, 238,
1909–1916 (2008).
Proceedings of ICAPP ‘10 San Diego, CA, USA, June 13-17, 2010
Paper 10141
11
16. Wierzbicki, B.W., Antal, S.P. and Podowski, M.Z.,
“On the Modeling of Gas Bubble Evolution and
Transport using Coupled Level-Set/CFD Method",
Nuclear Technology, 158, 2, 261-274 (2007).
17. Tselishcheva, E.A., Antal, S.P. and Podowski, M.Z.,
"Mechanistic Multidimensional Analysis of Horizontal
Two-phase Flows", Nuclear Engineering and Design
240, 405–415 (2010).
18. Tiwari, P., Antal, S.P. and Podowski, M.Z., “Three-
Dimensional Fluid Mechanics of Particulate Two-
Phase Flows in U-bend and Helical Conduits", Physics
of Fluids, 18, 4, 1-18 (2006).
19. Tiwari, P., Antal, S.P. and Podowski, M.Z., "Modeling
Shear-induced Diffusion Force in Particulate Flows",
Computers & Fluids, 38, 727–737 (2009).
20. Duderstadt, J., Hamilton, L., Nuclear reactor analysis,
John Wiley & Sons (1976).
21. Menikoff, R., Plohr., B., Rev. Mod. Phys., 61, 75
(1989).
22. Brackbill, J. U., D. B. Kothe and C. Zemach, "A
continuum method for modeling surface tension",
Journal of Computational Physics, 100, 2, 335-354
(1992).
23. Taylor, C. A., Hughes, T. J. R. and Zarins, C. K.,
"Finite element modeling of blood flow in arteries",
Computer Methods in Applied Mechanics and
Engineering, 158, 1-2, 155-196 (1998).
24. Franca, L. P. and S. L. Frey. "Stabilized Finite-Element
Methods II. The Incompressible Navier-Stokes
Equations", Computer Methods in Applied Mechanics
and Engineering, 99, 2-3, 209-233 (1992).
25. Sussman, M., E. Fatemi, P. Smereka and S. Osher.,
"An improved level set method for incompressible
two-phase flows", Journal of Computers & Fluids, 27,
Issue 5-6, 663-680 (1998).
26. Sussman, M. and E. Fatemi., "An efficient, interface-
preserving level set re-distancing algorithm and its
application to interfacial incompressible fluid flow",
Siam Journal on Scientific Computing, 20, 4, 1165-
1191 (1999).
27. Sussman, M., Almgren, A.S., Bell, J.B., Colella, P.,
Howell, L.H. and Welcome, M.L., "An adaptive level
set approach for incompressible two-phase flows",
Journal of Computational Physics, 148, 1, 81-124
(1999).
28. Sato, Y. and Sekoguchi, K., “Liquid Velocity
Distribution in Two-Phase Bubble Flow,” Int. J.
Multiphase Flow, 2 79-95 (1975).
Recommended